Відео

I love you.

00:00:03 - Intro & Motivation 00:01:06 - LLM & RL 00:08:36 - Challenges & Solutions 00:23:22 - Framework & Results

6:00 cannot agree more!

How to configure properly an on-premise S3 storage in Daft?

"give me fp16 or death" 😤😤

Hello snorkel team,i would love to be your part as Ml enginner,please let me know of further process

Good

“Laser”

clay shirky's Here Comes Everybody has come to life at AnyScale

讲得不错，给我很多启发

This guy is a head of his time - very sharp and makes a great point.

This guy is a legend. His work on RL especially TRPO, PPO algorithms is still the state of the art. Interesting interview 👍

Proud of MRNA tech and its utility? hmmmmmm

👑 👑 👑

I bought 12,000 shares of recursion pharmaceuticals

❤

Thanks for explaining this clearly.

Im glad Arthur Fleck really turn things around for him

Marc shouldn't talk authoritatively about military systems

Great content, horrible audio!

I'm somewhat struggling to understand the value of Ray Train over Lightning or even Vertex AI. It seems having to deploy onto a gke or eks cluster is more overhead than GCPs solution using on demand cluster spin up.

is this a toy example?

This guy is an affront to rational thought; a transparent self-aggrandizing shill not a fraction as intellectually capable as he believes himself to be by virtue of the simple fact that no one is or in practical terms could be

do people realize that Russia also has a drone program and by the end if the war it will be one if the most advanced

Good interviewer. Great convo

dr evil. plus conehead

It just seems all fake like scripted. You found a book on Singularity at a garage sale? 😂 You’re just that aligned, huh?

awesome!!!!

Great guy

Unfortunately VLLM Integration with KubeRay seems to be buggy for now. Couldn't get it to run, especially with the newer (0.6.4) VLLM versions. There only few resources and issues in the internet.

What? Last year's video?

Indeed Dr. Evil.

AWESOME

Awesome 👍

Requiring the companies to conduct and publish their safety research seems like a good regulation. If you don't want bad government regulation then maybe YOU need to have ideas for good regulations. Wait a minute! Are you waiting for Mitch McConnell and Pelosi to come up with something? They won't. But if your little toy robots go messing with the grid and so on, you can be sure the guys with the black SUV's will find you quick. They'll be angry. And they will have questions which you can't answer. You didn't study for the test? At all? And they'll be rude!

He’s obviously lying. A uh manoa student created ChatGPT, called Lana ai. Easy to google and fine that out.

Great discussion, love knowing more about how things evolved.

Fake it till you make it.

Interesting

Interesting

Perhaps the answer to scalable supervision is humans proficient at research given a RAG based system with academic libraries, the national archives, and high quality heterodoxy like Michael Hudson, John Stockwell, etc

This guy does not lack jaw power

Guy looks like Joker, now I'm really worried about AGI if anything bad happens to him.

♥♥♥♥I love AI♥♥♥♥

There should be much more investments in AI safety and much more effort put into AI safety regulations. Remember, when this goes very wrong there likely won't be a second chance for us. Very important warnings by two godfathers of AI, Geoffrey Hinton and Yoshua Bengio, and many other leading AI technicians, about a very real, possible AI existential threat! Like by creating subgoals when AIs are enabled to act autonomously, and one of those subgoals would be to get more control. When they'll surpass us in general intelligence, which we won't necessarily notice, we won't be able to keep up with them and imagine what they're planning or strategizing. Surely these people are ultimate generative AI technical experts and truly know what they're talking about. As a layperson I choose to trust their information and warnings before anyone else.

With the AI paradigm shift I guess people need to have a guarantee for a means of survival also without having a job, and that they have extra free time. Current intellectual and creative jobs also have boring repetitive time consuming tasks. AI is an opportunity and new way of life to continuously learn, explore and be more creative.

Wow, important conversation. Thanks for sharing! Great questions

Andreesen is brilliant

Will we all be substituted by super intelligent ai?